WhatsApp Group Integration
AI Agent Facilitating Group Planning With Friends
ROLE
Product Manager + Designer
Scope
Zero to One
Timeline
2 Months
A WhatsApp-based planning assistant designed to reduce decision fatigue in group chats. I led the product direction and MVP development through a pivot and two supporting MVPs. Each MVP offered unique insights that informed product strategy and shaped the path forward.
Project Scope
Over 10 weeks, I led the full product development cycle to address a familiar but underserved problem: making group plans actually happen in messaging apps like WhatsApp.
The project focused on building a conversational agent that could support the entire planning journey - from idea to execution - while working within real constraints of time and technical feasibility. The goal was to validate the concept through lean experimentation and define a solution that users could trust, understand, and actually use.
Frameworks Used
To bring Llama to life, I started by identifying a clear customer pain point through interviews and mapped out the MVP using the MoSCoW method to prioritize features. I defined the value proposition and success metrics through OKRs and KPIs, tested key flows via user testing, and iterated based on real feedback. Finally, I wrapped the process with a go-to-market strategy.
testing hypothesis
Rather than iterate linearly, I built two complementary MVPs in parallel to explore both functionality and user experience. One was a lightweight functional agent that mimicked interaction within WhatsApp; the other, a high-fidelity design prototype that visualized the complete planning flow.
Goal
Understand how users interact with a search agent embedded in WhatsApp and collect actual data on the type of queries.
Set Up
I created a business WhatsApp account and built an integration between WhatsApp and ChatGPT through Agentive, where I configured the agent's instruction base.
Limitations
✕ WhatsApp Business accounts can't be added to group chats, so testing was limited to 1:1 conversations.
✕ No visual UI, responses were purely text-based.
✕ Users couldn’t take action on core planning features like polls or availability pickers.
Testing Results
★ Top searches: restaurants, flights, coffee
★ 102 total messages exchanged
★ 2 to 5 messages per chat session
✕ 40% of users wanted increased speed
✕ 75% of users were not happy with the results
★ 100% of users said they would use it if results improved
This MVP didn’t solve the chaos in group planning but it gave users the feel of a search companion inside their most-used messaging app.
Goal
Limitations
User Stories
Testing Results
While it lacked functionality, the design validated that users valued visual clarity, quick decision tools, and having structure built into the chat itself.